Really, terrible, security vulnerability issues? DRI and networking kennel modules should absolutely not be able to interact with each other at all.

"kernel module" together with "should absolutely not be able to interact with each other" are an impossible requirement with Linux.

I think the other operating systems available for the Pi are roughly in the same boat (Windows & RiscOS). There was a nascent Minix port at some point, I wonder if it was abandoned.

Linux is (currently) a monolithic kernel and I'm not sure that can be accomplished without changing this.

The screen memory is taking up so much RAM that it's overlapping with regions of memory the network interface uses.

Resource are allocated via the kernel - it won't hand out overlapping address ranges.

Maybe the misbehaving driver is writing past the end of its requested space though, inadvertently? (I don't know if this is always called a "heap overflow" or if that's just Clang AddressSanitizer.)

Or something like https://mjg59.dreamwidth.org/11235.html is happening.

That resulted in a wide variety of different failures, from the kernel oopsing to various userspace components crashing. It would be very unusual to have unexpected DMA trigger such a specific failure.

(for avoidance of doubt, I wrote that blog post)

Out-of-bound memory write.

Why would that only interfere with the network driver, rather than tending to crash random userland or crash the kernel?

I don't know, let's see if anyone has an idea about it.

I was just explaining what the OP was asking for. I personally believe it's a EMI-related hardware issue.